DESIGN OF AN ULTRA-EFFICIENT GAN HIGH POWERAMPLIFIER FOR RADAR FRONT-ENDS USING ACTIVE

HARMONIC LOAD-PULL

Tushar Thrivikraman, James Hoffman

Jet Propulsion Laboratory, California Institute of Technology4800 Oak Grove Dr., Pasadena, CA 91109, USA

Tushar.Thrivikraman@jpl.nasa.gov

Keywords: GaN Power Amplifier, Active Load-pull,Harmonic tuning

AbstractThis work presents a new measurement technique, mixed-signal active harmonic load-pull (MSALP) developed byAnterverta-mw in partnership with Maury Microwave, thatallows for wide-band ultra-high efficiency amplifiers to bedesigned using GaN technology. An overview of the the-ory behind active load-pull is presented and why load-pullis important for high-power device characterization. In ad-dition, an example procedure is presented that outlines amethodology for amplifier design using this measurementsystem. Lastly, measured results of a 10W GaN amplifierare presented. This work aims to highlight the benefit of us-ing this sophisticated measurement systems for to optimizeamplifier design for real radar waveforms that in turn willsimplify implementation of space-based radar systems.

1 IntroductionGaN devices are increasingly becoming an enabling tech-nology for advanced space-based radar systems. The useof GaN allows for improved system performance due to itshigh efficiency and output power. In addition, the adoptionof GaN into commercial applications, such as cell phonebase stations has allowed the technology to mature and im-prove reliability and device performance.Such high priority missions as the proposed Earth RadarMission’s (ERM) DESDynI (Deformation, EcosystemStructure, and Dynamics of Ice) SAR Instrument (DSI) uti-lizing the SweepSAR concept are made feasible by the useof high power, high efficiency solid state power amplifiers[1, 2]. In the design concepts for this mission, multiple highpower TRM (transmit/receive modules) (> 100 W outputpower) that are thermally managed by the surface area ofthe TRM acting as the thermal radiator. For typical duty cy-cles and efficiencies, over 20 W of power must be dissipatedby this surface, requiring extremely large TRM, which arenot suitable for this space borne system.GaN technology presents two benefits for the proposedDSI: (1) a potential reduction of thermal stresses due to in-creased efficiency and (2) ability to operate at higher junc-tion temperatures. By easing the thermal requirement, theradiator size may be decreased allowing for a more com-pact TRM, reducing mission costs. In order to fully op-

timize the high power amplifier design using GaN HEMTtechnology, the new characterization technique, active har-monic load-pull, is being developed to increase efficienciesand reduce harmonic levels over a wide-band [3]. Thoughtuning load and harmonic impedances Class-J current andvoltage waveforms can be engineered that provide optimalbalance between performance, bandwidth, and linear re-sponse. There is increasingly more literature on the use ofthese systems for designing amplifiers for use with modu-lated telecommunications signals, but to the authors knowl-edge, no work has been demonstrated this technique fortypical radar signals.This work will present an overview of a wide-band mixedsignal active harmonic load-pull system (MSALPS) [4] thatis capable of harmonic tuning over a real radar waveform(such as a chirp) to optimize amplifier performance andefficiencies to > 70 %. This system was developed byAnteverta-mw in partnership with Maury Microwave anda detailed discussion of the load-pull system is presented in[5]. Section 1 will provide an overview the theory behindthe MSALPS system and the specific measurement setupused for amplifier characterization. Section 2 will present adesign overview of an L-band GaN amplifier for use in theproposed DSI TRM with the measured results of this designpresented in section 3.

2 Mixed-Signal Active Load-pullMeasurement Overview

The basic concept of “load-pull” is to present knownimpedances to devices under test (DUT) in order to de-termine their characteristic response, enabling performanceoptimization for non-linear devices. Fig. 1 shows the signalflow graph for a non-linear two-port device where a1,2 rep-resent the incident waves and b1,2 are the reflected waves.Sxx* represents the large signal s-parameters of the DUT.For a small-signal network analyzer, the DUT is insensitiveto load impedance and therefore standard VNA error cor-rection terms can be used to calibrate to the reference planeof the device to extract s-parameters. However, for largesignal conditions, load-pull is necessary since the DUTis no longer linear and therefore linear superposition andsimple two-port network theory is not valid. Impedancetuners allow a simulated ΓL presented to the device andenables the measurements of the non-linear large-signal S-parameters.

DUT

a1 b2

a2b1

S11*

S21*

S22*

S12*

Systemload

reflection𝛤L

Figure 1: Signal flow diagram for a large-signal load-pullmeasurement system.

Traditionally, load-pull systems are passive, using mechan-ical probes to create a mismatch that presents the desiredimpedance by reflecting signal back to the DUT. Thesesystem have been used for years for amplifier develop-ment with modern tuners being electronically controllableand able to tune fundamental and harmonic impedances.Passive systems however are constrained due to tuner andsystem losses that limit the maximum reflection coeffi-cient achievable. For typical high power devices, outputimpedances are below 5 Ω, and therefore are difficult tocharacterize on passive load-pull systems. Another limita-tion for wide-band signal are the time-consuming tuner cal-ibrations that are only valid to for single-tones and/or har-monics requiring characterization at different fundamentaltones within the frequency band of interest to optimize theamplifier design for wideband operation.Many of these performance limitations of passive tuner sys-tems can be addressed by active load-pull. As the namesuggests, active load-pull is achieved by injecting a signalinto the DUT to create the desired impedance. The ac-tive injection can be performed by either closed- or open-loop methods. The closed-loop method, Fig. 2a, injects asampled output signal back into the DUT with appropriatephase and amplitude to produce the desired reflection. Thismethod optimal for fast device characterization, but can beprone to oscillations and requiring filtering. The MSALPSthat is discussed in this work is an open-loop active system(Fig. 2b), that utilizes independent arbitrary waveform gen-erators to create independent channels for source and loadfundamentals and harmonics.An open-loop system requires more signal processing andsoftware iterations since the reflected wave is generated in-dependently from the incident wave. However, the addedcomplexity allows flexibility with the ability to tune wide-band modulated signals and have independent control ofharmonics as well as the ability to achieve very high reflec-tion coefficients as they are only limited by the power sup-plied by the injection amplifier. Fig. 3 shows the MSALPSconfigured for four tuning loops. Anteverta-mw and MauryMicrowave’s implementation utilizes a National InstrumentPXI chassis, which allows rapid sampling of the incidentand reflected waves. Each tuning loop requires a separatedigital AWG that generates the appropriate waveforms andis unconverted to RF for injection into the device. This pro-cedure is handled automatically through the software andalgorithms developed by Anteverta-mw. The directionalcouplers are used to measure the a and b waves from theinput and output and digitized for software analysis. A

ɸDUT

injection amplifierbias tee amplitude and phase

controlcouplerRF source

DUT

injection amplifierbias teeAWG source AWG source

(a)

(b)

Figure 2: Closed (a) and open (b) active load pull systemconfigurations. In (a) for the closed loop case, the injectedsignal is a sampled version of the output signal while in theopen loop case (b) the injected signal is generated by anindependent source.

diplexer is used to power combine the fundamental and har-monic signals for injection into the device. Tuning loop Ais used for the fundamental source, while tuning loops B, C,and D are used for the load fundamental, second, and thirdharmonic respectively.For these measurements, the load fundamental, second,and third harmonics are swept to determine optimumimpedance. The source fundamental can also be swept withthe match condition as close to the small-signal S11 conju-gate match for optimum power transfer. In addition, a sta-ble phase reference is used to determine the dynamic I-Vcharacteristics of the device, which can be used to optimizethe waveform for Class-J operation and more sophisticatedwaveform engineering.The measurement is performed in two steps. The first stepis to determine the proper injection signals for all powersand impedance conditions. This step is achieved throughthe use of the AWGs and adjusting the waveforms on apulse-by-pulse basis to achieve the necessary powers andimpedances at both the source and load fundamentals andharmonics. The second step of the measurement routinethese calculated injection signals are used to fully character-ize the device with both RF and DC parameters measured.Detailed measurement theory of the MSALPS can be foundin [6, 7, 8].Another important feature of the MSALPS is the ability toload-pull modulated signals that have both frequency andamplitude variation. This allows for very precise amplifiercharacterization and tuning to optimize the performance forspecific radar waveforms. The requirements for the pro-posed DSI to achieve the desired resolution require chirpbandwidths as wide as 80 MHz at L-band. Designing opti-mal amplifier performance over this band is extremely chal-lenging with conventional load-pull systems. Fig. 4 is anexample linear chirp waveform that can be programmedinto the MSALPS system.

3 Amplifier Design OverviewThis section outlines the basic steps to use the MSALPS todesign a highly efficient HPA. The steps are as follows:

DigitalAWG

x4DigitalA/D

PA @ fo

Output section

b2,f0 b2,2f0 a2,f0

LO LO

Input section

To DC

Bias Tee

High power fixture with bias decoupling

I 1

On-wafer configurationTo DC

To DC

BaseBand I V

To DC

Digital AWG Digital A/D Digital AWG

PA @ 2f0

PA @ f0RF @ f0

RF @ 2f0

I/Q f0 signal

I/Q 2f0 signal

PA @ 2f0

PA @ f0 RF @ f0

RF @ 2f0

I/Q f0 signal

I/Q 2f0 signal

DUT

DUT

a1,f0 a1,2f0 b1,f0 b1,2f0

LO LO LO

a2,2f0

LO

a1,f0 a1,2f0 a1,BB b1,f0 b1,2f0 b1,BB b2,f0 b2,2f0 b2,BB a2,f0 a2,2f0 a2,BB

RF Source

LO Source

To LO

X2

To RF@ f0

To RF@ 2f0

HPR

aREF

BaseBand I V

Bias Tee

I V BaseBandReference

Planes

I V BaseBand

Reference Planes

Load Reference

Plane

DUT Output section

b2,f0 b2,2f0 a2,f0

LO LO

Input section

To DC

Bias Tee

High power fixture with bias decoupling

I 1

On-wafer configurationTo DC

To DC

BaseBand I V

To DC

Digital AWG Digital A/D Digital AWG

PA @ 2f0

PA @ f0RF @ f0

RF @ 2f0

I/Q f0 signal

I/Q 2f0 signal

PA @ 2f0

PA @ f0 RF @ f0

RF @ 2f0

I/Q f0 signal

I/Q 2f0 signal

DUT

DUT

a1,f0 a1,2f0 b1,f0 b1,2f0

LO LO LO

a2,2f0

LO

a1,f0 a1,2f0 a1,BB b1,f0 b1,2f0 b1,BB b2,f0 b2,2f0 b2,BB a2,f0 a2,2f0 a2,BB

RF Source

LO Source

To LO

X2

To RF@ f0

To RF@ 2f0

HPR

aREF

BaseBand I V

Bias Tee

I V BaseBandReference

Planes

I V BaseBand

Reference Planes

Load Reference

Plane

a1 b1

NI PXI Chassis

b2 a2

PA @ fo

PA @ 2fo,3fo

source fo

load foload 2fo

load 3fo

receiver receiver receiver receiver

Figure 3: Block diagram of mixed-signal active load-pull system (MSALPS). The system is configured for four tuningloops, the fundamental source, and load fundamental, second, and third harmonics. Each of the loops require an individualarbitrary waveform generator that produces the appropriate waveform to inject into the device to produce the desired Γ.

(a) (b)

Figure 4: Modulated chirp waveform in the (a) I-Q timeand (b) frequency domains. The modulated waveform canbe used to synthesize actual expect input waveforms, im-proving system design fidelity.

1. Perform DC output characteristic sweep to determineappropriate quiescent point of operation. Class-J am-plifiers are typically biased near class-B operation.

2. Measure small-signal S-parameters of device to deter-mine input and output matching conditions.

3. Perform fundamental source and load-pull to deter-mine optimal matching regions for given bias points

4. Perform harmonic tuning to adjust second and/or thirdharmonic to present unmatched conditions at those fre-quencies

An example load-pull sweep configuration from step 4 isshown in Fig. 5, which shows the impedance points thatare used to tune the source fundamental, load fundamental,and load harmonics. The sweep was performed from linearto saturated power at 10 % duty cycle.For reference, a sample single-tone measurement result ispresented in this abstract. Here, a 10 W GaN HEMT de-vice is characterized in a high power test fixture, pictured

!

source fo tuning

load fo tuning

load 2 & 3 fo tuning

Figure 5: Typical load-pull point constellation for tuningsource fundamental load fundamental and second harmon-ics. The source and load fundamental are centered aroundthe small signal conjugate reflection coefficients, while thesecond harmonic is tuned for maximum reflection.

in Fig. 6 with the bias tees. The device was biased nearclass B operation since typical high power radar amplifiersare operated in saturation, linearity is not a major concernfor radar HPAs. The measurement was calibrated to thedevice reference plane, but still includes package parasiticlosses.For the initial design, optimal source and load impedancesat the fundamental were tuned for the 10 W GaN HEMTdevice biased in class B operation. The source match wasmatched to the large-signal Γin while the output was sweptaround the small-signal S22 conjugate match. Output powerof 40 dBm and 68 % PAE are shown in Fig. 7 without the

deembeded reference plane

Figure 6: 10W GaN device in test fixture for load-pullshowing reference plane with fixture deembeded.

−0.6 −0.5 −0.4 −0.3 −0.2 −0.1 0 0.1

−0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

PAE (%)Pout fo (dBm)Pout 2fo (dBm)Pout 3fo (dBm)

64

68

higherPAE

41

40

39

higherpower

1820 lower

2fo22

20lower3fo 22

26

24

Figure 7: Design of 10W GaN HEMT device in class Boperation with no harmonic control. Output power of of 40dBm approaching 70% PAE are achieved.

use of active harmonic control. Even though this perfor-mance is acceptable, the second and third harmonic com-ponents are high (>20 dBm). The contours for PAE andharmonics show a tradeoff between minimizing either sec-ond or third harmonic content and optimal efficiency. Forexample, by choosing a load impedance to reduce the sec-ond harmonic, the third harmonic becomes larger, tradingoff efficiency. Using the MSALPS for harmonic tuningadds the additional control knobs to optimize impedancesindependently for both the second and third harmonic, thusimproving the overall efficiency.These results can be directly compared to Fig. 9, wherethe additional two tuning loops of the MSALPS sys-tem were used to control the second and third harmonicimpedances. In order to achieve maximum efficiency a re-flective impedance (open or short) is used.For this sweep, the source and load second and third har-monic were tuned to optimal points within the given tun-ing range as presented in Fig. 5, converging on the finalimpedance matching values shown in Fig. 8. Sweeping thefundamental load, the PAE and power contours are shownin Fig. 9. By controlling the harmonic content, the effi-ciency can be improved to over 76% while still achieving

!Lfo=#$0.3#+#0.3#j

!Sfo=#$0.84#+#0.12#j

!L3fo=#$0.5#+#0.8#j!L2fo=#0.21#+#0.9#j

Figure 8: Final optimal matching impedances showing fun-damental source, load, and harmonic impedances.

−0.5 −0.4 −0.3 −0.2 −0.1 0

0

0.1

0.2

0.3

0.4

0.5

PAE (%)Pout fo (dBm)Pout 2fo (dBm)Pout 3fo (dBm)

74

76

higherPAE

41

40

39

higherpower

47

lower 2fo

10

5lower3fo 8

11

Figure 9: PAE and output power contours for 10W GaNHEMT device with a single tone stimulus. The PAE isshown in magenta, with the fundamental, second, and thirdharmonic shown in red, blue, and green respectively.

over 10 W output power, and reduced harmonic output. Theoptimal point for PAE still highlights the tradeoff betweenpower at 2fo and 3fo. This technique can also be used toreduce harmonic content to comply with international radioemission guidelines.To further gain insight of the amplifier characteristicsthe dynamic I-V waveforms can be calculated using theMSALP software. In order to fully understand the currentand voltage waveforms within the device as presented in[9], the package parasitics and internal drain-source capac-itance should be deembeded to the internal current gener-ator of the transistor as shown in Fig. 10. Using a com-mercial simulator tool, the fixture, package, and internaldevice CDS were modeled and extracted with the match-ing impedances translated to the internal device referenceplane as shown in Fig. 11.Understanding this transformation from the package refer-ence plane to internal current generator is important for har-monically tuned amplifiers. The fundamental and harmonicimpedances must be controlled over a wide bandwidth tomaximize performance. As shown in Fig. 11 the parasiticshave a large effect on the matching conditions at the internalgenerator.

fixture

package

Cds

device

intrinsic device

RL

Figure 10: Model of package and device parasitics deembe-ded to allow measurement of dynamic current and voltageat the intrinsic device current generator.

Figure 11: Shifted matching impedances to present to in-trinsic device removing effects of package parasitics anddrain source capacitance.

Once these parasitics are removed and the impedance pointsare shifted to the internal generator reference plane, dy-namic current and voltage waveforms can be calculated.Fig. 12 plots the current and voltage versus time at 3 dBcompression, showing the peak voltage approximately at 55V, with a peak current of 1.4 A. Fig. 12 plots the complexload line for for the matched amplifier. This load line in-dicates class B like waveforms due to the ideal circular na-ture of the response. The negative portions of voltage andcurrent waveforms indicate that there are additional para-sitics capacitances that were not successfully deembeded,and therefore further optimization of the package model isrequired.

4 SummaryIn summary, this abstract presents an overview of mixed-signal active load pull system that can be used for devel-oping RF front-ends for space-based radars. The load-pullsystem may be used to improve amplifier efficiencies overa wide-bandwidth, reducing thermal constraints and space-craft bus requirements for such missions as the proposedDSI.A 10 W GaN amplifier was designed to achieve over 76 %

.0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6

−10

0

10

20

30

40

50

60

time (ns)

Dra

in V

olta

ge (V

)

−0.35

−0.1

0.15

0.4

0.65

0.9

1.15

1.4

Dra

in C

urre

nt (A

)

Figure 12: Current and voltage waveform at intrinsic de-vice, showing peak currents and voltages.

−10 0 10 20 30 40 50 60−0.5

0

0.5

1

1.5

Voltage (V)

Cur

rent

(A)

Figure 13: Complex load line for matching amplifier at 3dB compression.

efficiency by optimizing the harmonic content. In addition,dynamic current and voltage waveforms were calculated atthe intrinsic device current generator by removed packag-ing effects and drain source capacitance. The purpose ofthis work is to highlight the general procedure and achiev-able performance of developing high power GaN amplifiersusing this technique and the benefit to the instrument de-sign through optimization. Future work includes develop-ing matching networks to achieve these impedances as wellas characterize the amplifier using a modulated response.

AcknowledgmentThis work is supported by the NASA Earth Radar Missiontask at the Jet Propulsion Laboratory, California Institute ofTechnology, under a contract with the National Aeronauticsand Space Administration. c©2012 California Institute ofTechnology. Government sponsorship acknowledged.

References[1] P. Rosen, H. Eisen, Y. Shen, S. Hensley, S. Shaffer,

L. Veilleux, R. Dubayah, K. Ranson, A. Dress, J. Blair,S. Luthcke, B. Hager, and I. Joughin, “The proposed

DESDynI mission - from science to implementation,”in IEEE Radar Conference, pp. 1129 –1131, May 2011.

[2] J. Hoffman, L. Del Castillo, D. Hunter, and J. Miller,“Robust, rework-able thermal electronic packaging:Applications in high power tr modules for space,” IEEEAerospace Conference, 2012.

[3] P. Wright, J. Lees, J. Benedikt, P. Tasker, and S. Cripps,“A methodology for realizing high efficiency class-J ina linear and broadband PA,” IEEE Transactions on Mi-crowave Theory and Techniques, vol. 57, pp. 3196 –3204, Dec. 2009.

[4] S. Dudkiewicz and M. Marchetti, “Active load pull sur-passes 500 watts!,” tech. rep., Maury Microwave andAnteverta-mw, Nov. 2011.

[5] M. Marchetti, “Mixed-signal active load pull: The fasttrack to 3g and 4g amplifiers,” MICROWAVE JOUR-NAL, Sept 2010.

[6] M. Marchetti, M. Pelk, K. Buisman, W. Neo, M. Spir-ito, and L. de Vreede, “Active harmonic load-pull withrealistic wideband communications signals,” IEEETransactions on Microwave Theory and Techniques,vol. 56, pp. 2979 –2988, Dec. 2008.

[7] M. Squillante, M. Marchetti, M. Spirito, andL. de Vreede, “A mixed-signal approach for high-speedfully controlled multidimensional load-pull parameterssweep,” in ARFTG Microwave Measurement Confer-ence, pp. 1 –5, june 2009.

[8] M. Marchetti, R. Heeres, M. Squillante, M. Pelk,M. Spirito, and L. de Vreede, “A mixed-signal load-pull system for base-station applications,” in IEEE Ra-dio Frequency Integrated Circuits Symposium, pp. 491–494, may 2010.

[9] S. C. Cripps, RF Power Amplifiers for Wireless Com-munications, Second Edition (Artech House MicrowaveLibrary (Hardcover)). Norwood, MA, USA: ArtechHouse, Inc., 2006.

Tushar Thrivikraman is currently an RF engineer at theJet Propulsion Lab at the California Institute of Technologyin the Radar Science and Engineering section where he hasfocused on the development of RF hardware for air- andspace-borne SAR imaging systems. He received his PhD inElectrical and Computer Engineering from the Georgia In-stitute of Technology in 2010. His research under Dr. JohnCressler in the SiGe Devices and Circuits Research Groupat Georgia Tech focused on SiGe BiCMOS radar front-endsfor extreme environment applications.James Hoffman is a Senior Engineer in the Radar Tech-nology Development Group at JPL. He received his BSEEfrom the University of Buffalo, followed by MSEE and PhDfrom Georgia Tech in planetary remote sensing. He hasworked in the design of instruments for remote sensing ap-plications for more than 10 years. In previous technologydevelopment tasks, he successfully developed a new low

power digital chirp generator, which has been integratedinto several radar flight instruments. He has experience de-signing radar systems for both technology development andspace flight hardware development, and is currently the RFlead for the proposed DESDynI radar instrument.